At Google I/O 2026, Sundar Pichai unveiled Gemini 3.5 Flash, a 24/7 personal agent, and $180-190B in capex to back the company's agentic AI ambitions.
Google spent its biggest developer conference of the year making a single argument: artificial intelligence should act, not just answer. At I/O 2026, held May 19 and 20 in Mountain View, CEO Sundar Pichai called it the start of the agentic Gemini era. The word agent appeared alongside nearly every product shown.
Usage numbers back the ambition. The Gemini app has crossed 900 million monthly users, more than doubling the 400 million reported a year ago, with daily requests growing roughly sevenfold. Model APIs now process about 19 billion tokens per minute, Pichai said on stage.
The headline launch was Gemini 3.5 Flash, the first model in the 3.5 family, which became generally available on opening day across the Gemini app, AI Mode in Search, the Gemini API, the Enterprise Agent Platform, and Antigravity, Google's developer environment. According to Forbes, it scores 76.2% on Terminal-Bench 2.1 and 83.6% on MCP Atlas, and runs roughly four times faster than comparable frontier models. A larger Gemini 3.5 Pro is already in internal use and expected to ship next month.
Beyond the phone
Google previewed Gemini-powered features for Android Auto and Google Built-in, extending the model into in-car experiences. As Mashable reported, the updates are designed to help drivers with navigation, music, and conversation while keeping attention on the road. The conference also featured a 24/7 personal agent and an AI-powered shopping assistant. Google called AI Mode in Search the deepest change to that product in its history, without specifying the impact on advertising or organic ranking.
The infrastructure bet
Pichai told the audience to expect capital expenditure of $180 billion to $190 billion this year, up sharply from $31 billion in 2022. Most of the investment targets the eighth-generation TPU 8t and TPU 8i chips for training and inference. Google also introduced Gemini Omni, a separate multimodal model, with limited detail at launch.
Who else is moving
The reframing of Gemini as an agent platform reflects an industry-wide pivot. Benchmark performance now takes second place to whether a model can complete tasks reliably without constant supervision. BleepingComputer reported that OpenAI, in strategy documents surfaced during the U.S. v. Google antitrust case, has signaled plans to push ChatGPT beyond apps and websites into new form factors by 2026, with specifics still redacted from public filings.
Enterprise adoption is accelerating in specialized domains. Bristol Myers Squibb announced this week a partnership with Anthropic to deploy Claude to more than 30,000 employees, targeting drug discovery and clinical research. The deal illustrates how artificial intelligence in medicine is moving from pilot to production, and McKinsey estimates that agentic AI could lift clinical development productivity by 35% to 45% over five years, according to The News.
Nine hundred million monthly Gemini users gives Google a distribution advantage most competitors cannot replicate. That scale, however, also raises the stakes: agentic tasks are less forgiving than chatbot exchanges. A wrong answer is annoying. A wrong action can cause real damage. Reliability, not speed metrics alone, will determine whether enterprises commit at scale.
Google has made its bet in capital. Whether the models deliver at production quality on Pichai's timeline is what developers and enterprise buyers will be testing over the next several quarters.
---
FAQ
What is Gemini 3.5 Flash?
Gemini 3.5 Flash is the first model in Google's Gemini 3.5 family, released publicly at I/O 2026. Google says it outperforms Gemini 3.1 Pro on coding and agentic benchmarks and runs roughly four times faster than comparable frontier models at similar capability.
How many users does the Gemini app have?
Sundar Pichai reported at I/O 2026 that the Gemini app has crossed 900 million monthly users, up from 400 million a year ago, with daily requests growing approximately sevenfold in the same period.
How much is Google investing in AI infrastructure in 2026?
Google expects capital expenditure of $180 billion to $190 billion this year, up from $31 billion in 2022, with a large share directed at eighth-generation TPU chips designed for AI training and inference workloads.
Will Gemini be available in cars?
Yes. Google previewed Gemini integration for Android Auto and Google Built-in at I/O 2026, with features designed to assist drivers on navigation, music, and conversation without requiring attention away from the road.